AI300
  1. Sebastien Bubeck / OpenAI MTS
    1801.
    I feel the First Proof results are a bit downplayed ...›
    LLM score 92 · 5 months ago
  2. Andrew Lampinen / Research Scientist at DeepMind
    1802.
    Really cool work — learning over sequential experiences that contain the embodied cue of viewpoint as well as visual inputs, can give rise to human-like 3D shape perception! https://t.co/3rfigNmAzq
    LLM score 75 · 5 months ago
  3. Sholto Douglas / Researcher at Anthropic
    1803.
    Reiner taught me much of what I know - goes without saying that I trust him to make the best chip in the world.›
    LLM score 75 · 5 months ago
  4. Andrej Karpathy / AI researcher
    1804.
    CLIs are super exciting precisely because they are a "legacy" technology, which means AI agents can natively and easily use them, combine them, interact with them via the entire terminal toolkit.›
    LLM score 92 · 5 months ago
  5. Jim Fan / NVIDIA Director of Robotics
    1805.
    What can half of GPT-1 do? We trained a 42M transformer called SONIC to control the body of a humanoid robot.›
    LLM score 92 · 5 months ago
  6. Cameron Wolfe / Researcher at Netflix
    1806.
    Interesting study on properly using PDF data with LLMs.›
    LLM score 92 · 5 months ago
  7. Merve Noyan / Hugging Face ML Engineer
    1807.
    everyone talks about how perplexity AI is dead? they actually released four SOTA diffusion-based embedding models just last week in 600M and 4B 🔥›
    LLM score 92 · 5 months ago
  8. Devansh Pandey / Standard Intelligence Cofounder
    1808.
    shout out to the trusty heap for storing all the video data needed to train FDM-1.›
    LLM score 82 · 5 months ago
  9. Lucas Beyer / Meta Researcher
    1809.
    New optimizer with earth-shattering plots making the rounds, and published in Nature too (Machine Intelligence, but let's just drop that part.)›
    LLM score 92 · 5 months ago
  10. Merve Noyan / Hugging Face ML Engineer
    1810.
    I didn't want to comment on this but I saw Anthropic employees who mentioned distilling one model isn't innovation, just read papers from these labs they have good stuff there? ›
    LLM score 92 · 5 months ago
  11. Chace Lee / ex MTS at xAI
    1811.
    Awesome work. VPT is one of my favorite papers. Computer-use agents live or die by where you draw the abstraction boundary.›
    LLM score 92 · 5 months ago
  12. Cameron Wolfe / Researcher at Netflix
    1812.
    Rubric-based RL is a really popular topic right now, but it’s not new.›
    LLM score 85 · 5 months ago
  13. Christian Szegedy / ex xAI Cofounder
    1813.
    Just a different way of saying: AI is getting increasingly good at long-horizon tasks.›
    LLM score 75 · 5 months ago
  14. Vahid Kazemi / ex MTS at xAI
    1814.
    AI-generated music went from 'AI slop' to 'this is as good a song as I’ve ever heard' real quick.
    LLM score 72 · 5 months ago
  15. Devansh Pandey / Standard Intelligence Cofounder
    1815.
    there is much work to do. in the next few months we need to train 100x larger models on 200x as much data as we have done before, scale RL to have millions of agents learning things from the internet, and use the computer use prior to build the first models that can scalably https://t.co/vzBqlj3kBi
    LLM score 92 · 5 months ago
  16. Damek Davis / Assoc. Professor Wharton Stats
    1816.
    Very cool! If I had time, I would try to get low parameter count by taking, e.g., the 512 parameter model and then distilling it down to a smaller model.›
    LLM score 75 · 5 months ago
  17. Jeff Dean / Chief Scientist at DeepMind
    1817.
    AI in educational settings has enormous positive potential (education is one of seven areas I & co-authors discussed in our recent paper on Shaping AI's Impact on Billions of Lives https://t.co/tjHmjn2vwj).›
    LLM score 92 · 5 months ago
  18. Noam Brown / OpenAI Research Scientist
    1818.
    tl;dr SWE-bench Verified is heavily contaminated for all frontier models, and many of the problems are also broken.›
    LLM score 95 · 5 months ago
  19. Aidan Smith / Cofounder of Flapping Airplanes
    1819.
    if you’ve ever asked yourself if like, 5 guys could’ve invented and scaled gpt, from the data to the arch to the posttraining, you’ve got your answer.›
    LLM score 92 · 5 months ago
  20. John Carmack
    1820.
    I always lost performance when I tried to use silu/gelu activations in my RL value networks, and I finally understand why.›
    LLM score 98 · 5 months ago
  21. Jack Clark / Anthropic Cofounder
    1821.
    Figuring out what the trends will be for AI and employment feels like figuring out how deep learning might influence computer vision in ~2010 - clearly, something significant will happen, but there is very little data out of which you can make a trend.
    LLM score 92 · 5 months ago
  22. Merve Noyan / Hugging Face ML Engineer
    1822.
    OlmOCR-Bench by @allen_ai is now a @huggingface Benchmark dataset 🏆›
    LLM score 75 · 5 months ago
  23. Ronak Malde / DeepMind Researcher
    1823.
    Been having several convos lately with people asking "are all software engineers going to get cooked in the next few years?".›
    LLM score 85 · 5 months ago
  24. Christian Szegedy / ex xAI Cofounder
    1824.
    A fair assessment. It looks like the gap in our timelines is narrowing.›
    LLM score 92 · 5 months ago
  25. Chace Lee / ex MTS at xAI
    1825.
    Ideas are cheap. Execution scales.›
    LLM score 75 · 5 months ago
  26. Ben Burtenshaw / Hugging Face Researcher
    1826.
    this > llama-server > run my vanilla python harness›
    LLM score 75 · 5 months ago
  27. Igor Babuschkin / Cofounder of xAI
    1827.
    Building great AI products requires excellence in both creativity and technical execution.›
    LLM score 85 · 5 months ago
  28. Christian Szegedy / ex xAI Cofounder
    1828.
    It makes me think: would it make sense to build larger/more capable models from slightly modulated smaller models as components? (Eg.›
    LLM score 82 · 5 months ago
  29. Jack Clark / Anthropic Cofounder
    1829.
    Choose your fighter. From a paper I'm writing up for Import AI this week about the behavior of language models in a simulated nuclear crises.›
    LLM score 92 · 5 months ago
  30. Cameron Wolfe / Researcher at Netflix
    1830.
    Really interesting reward modeling approach that works well for small LLM judges: https://t.co/7Wqd9tiUsN›
    LLM score 85 · 5 months ago
More